Visualization vs. Analytics Part III – Analytics Tools & Conclusion
The Workings of Analytics Tools
Analytics tools from Aurora Predictions, Alteryx, BigML, and Datameer to name a very few are a new class of enterprise tools that apply mathematics and statistics on historical multi-dimensional and multi-source data to make forecasts and predictions. These tools are also AI enabled to better drive strategy and decisions.
AI, as contained in most analytics tools, is in the form of Machine Learning (ML) that has many powerful uses. The figure to the right from BigML1, shows an application of the ML algorithm of Decision Tree, which is used to target individuals of interest with income over $50K.
Other examples of Decision Tree use include finding the customerswho will proceed to checkout, what are the likely modes of failure in the supply chain, which drugs are most likely to proceed to the next phase of trials, etc. However, as seen on the figure, ML is not intuitive, in fact, it’s technical, statistical, and requires data science expertise.
The table below from the Finance Analytics Institute2 highlights the complexity of ML through the progression of its areas of ML Process, ML Division, ML Learning, and ML Algorithm. Each area sequentially combines to deliver an ML result.
The ML Process starts with the problem (DefineObject) to be solved, which is often done by a consultant. Once defined, data is collected, cleaned, and prepared. Then data scientists determine the ML Division, ML Learning and ML Algorithm to program the model for the problem to be solved.
The model is next “trained” and tested to determine if training was successful. If all goes well then, the model is ready for deployment, but at any point in the ML Process, a failure can send the process back to any intermediate step or the beginning.
With this complexity, it is no wonder ML projects often hit a wall3 as “eight out of ten organizations reported their AI and ML efforts have stalled” and “96% of companies have experienced problems with data quality . . . required to train AI and build model confidence”.
To properly understand analytics tools, it is first necessary to make the distinction between Analytics, AI, and ML.Analytics is about the application of mathematics and statistics. Mathematics includes the disciplines of calculus, algebra, geometry, trigonometry; whereas mathematical statistics is about probability and outcomes. Note, that Analysis, when used in the context of reporting and visualization, typically refers to the Arithmetic functions to add, subtract, multiply, and divide.
Artificial Intelligence (AI) at the high level is a machine making a human decision. Any technology that can create this action is considered a form of AI. ML is one such technology, and therefore a subset of AI. So too are technologies like Systematic Intelligence™, neural networks, etc.
The wondrous image to the right from NASA4 of a black hole is created by ML. We can’t be sure this image is correct because we’ve never actually seen a black hole, but by combining astronomer’s data and algorithms from across the globe, the best unbiased image was rendered.
Analytics tools derive their power from mathematics, statistics, and AI; whereas, visualization is largely filled with arithmetic.There are many very powerful down- to-earth applications of analytics and AI in unbiased forecasting and predictions.
For the applications of analytics tools, we return to the five questions for data driven decisions of: What Happened, Where it Happened, Why it Happened, What Will Happen, and How To Make It Happen? The wheelhouse for analytics tools is the last three questions of Why, What Will, and How To.
Decisions are all about the future! Consider, have you made a decision about the past? No, because the past has happened. All that can be said about a decision made in the past is whether that decision had the desired outcome of its future. However, we do use the past to make predictions about the future.
Analytics enables quantitative and unbiased forecasts and predictions about the future using data of the past. This, as compared to the biased forecasts that are typically done; i.e. forecasts made by humans.
While humans have a good intuition of the future and may have specific knowledge of it as well, predictions are still at the mercy of human needs and emotions; e.g. I need to have this deal close this quarter, my boss wants more revenue, if we don’t make budget someone will take the blame, etc. Alternatively, analytics have no qualms with pride, prejudice, or politics so these calculations cater to objectivity.
Further, humans are not inherently quantitative creatures. We naturally learn language but require much effort to learn mathematics. This inhibits what we do analytically to mostly small and high-level computations.
Conversely, analytics tools engage the mathematical and the detailed to enable us to do that which we cannot through our limited tools of spreadsheets, BI, and visualization.
Applications of analytics and AI are vast. At the Analytics Academy there is a session on Applications of Advanced Analytics with examples that include: “Fair Challenge” to assign stretch revenue goals that produce a similar probability for all parties to achieve, AI forecasting for more accurate long-range planning, Monte Carlo Simulation for inventory optimization, correlations to foot traffic in retail stores to know when and where to spend promotion capital, and correlations of leading economic indicators to learn what affects demand.
So too, analytics are deployed for trend prediction as seen on the image to the right from Aurora Predictions and as reviewed at the Analytics Academy. Here we have a manufacturer of business and consumer electronics that are distributed by a variety of retailer (city) stores through the U.S. The box at the right- hand side of the screen is the Drill Path that systematically drills through data.
The displayed Drill Path predictively finds those distributors, stores, and products that have good trends YTD but predicted to go bad. The report in the middle of the image, automatically renders the results to display the numerical YTD trend and an arrow to indicate the future prediction of the trend. Here we find the “horses that will leave the barn” in advance so we can “close the barn door” before they leave.
The table to the right displays some of the many and more varied applications of analytics throughout a company.
Analytics tools provide the highest value to business because it can deliver unbiased predictions of the future, and with that, the ability to make better and strategic data driven decisions.
If spreadsheets and visualization are all that is in the Finance and Operations toolbox, then the best that can be achieved is some influence on tactical decisions as a partial 2nd Generation Partner. To go the full 10 yards, analytics are essential to become the strategic 3rd Generation Partner – The Analytics Business Partner.
As seen in the D&B/Forbes5 figure below, setting budgets aside, the challenges of analytics are about complexity, data quality, culture, alignment, and skills gap (knowledge of mathematics). As such, the adoption of analytics tools is materially higher than data visualization. Further, the nature of many graphical outputs (e.g. Decision Tree) make it difficult for executives to understand, and therefore they tend to discount what they can’t comprehend.
Complexity requires specialized skills. For example, ML has numerous forms (e.g. R and Python) and many differenteditors (even in the same form) – which to use? Further, the ML, statistical, and mathematical libraries in analytics toolscan be large, and you’ll need to know what the formulas do in order to properly apply them;
e.g. can I use the SPCI for fraud detection, when is Bayesian inferential modeling better than referential correlations, can Decision Tree give a better prediction than Support Vector Machine, etc.
Enterprise analytics vendors take effort to make their tools more usable, but there remains an underlying design view that users will be more like data scientists than Excel analysts. Some vendors like Alteryx position their product as “self- serve analytics” or Aurora Predictions that is “Snap-on analytics and AI for Excel Users”. Both purports to be for the business analyst, but only Aurora Predictions has AI without ML and does not require the user to know statistics, thus enabling all users access to AI forecasting and predictions.
Also, we distinguish enterprise analytics tools from desktop statistical tools and analytics cloud platforms. Desktop stat tools (e.g. Crystal Ball, Mini Tab, etc.) are basically spreadsheet extensions. Used for small data sets, these tools typicallydon’t have AI and require knowledge of the application of mathematics and statistics.
Analytics cloud platforms (e.g. Microsoft Azure, AWS, Oracle DataScience, etc.) are for use with large multi- source data and are the purview of consultants, data scientists, programmers, and application developers.
As such, when engaging analytics tools, the user skills will be determining; i.e. the typical business analysts is an Exceljockey but not a statistical/mathematical scholar. Also, analytics tools can require new workflows because most need a specialized skill set.
Therefore, the less mathematical, data science, programming, and database knowledge needed by a business analyst to use an analytics tool, the better the tool can be used, the faster it will be adopted, the less intrusion into the existing workflow, and the quicker to deliver benefits from insights and foresight.
To reach the plateau of Analytics Business Partner, it’s about bringing the right tool to the right job. You can’t build a house with only a hammer, and you can’t gain insights and foresights from only data visualization tools.
Can we give them the instruction manual on how to bring the right tool to the right job?
The value of having both tool sets is to answer the five key questions to making data driven decisions for business growth and optimization of: What Happened, Where it Happened, Why it Happened, What Will Happen, and How To Make It Happen?
As seen on the image to the right from D&B/Forbes1, organizations have a toolbox to do ”analytics” that they lump together to include spreadsheets, dashboards, and true analytics.
It’s not one tool before the other, or one instead of the other. They are both needed and provide different values; data visualization informs about the past, while analytics reveals the current and future.
Note, while analytics tools overlap many visualization capabilities, data visualization is better suited for key operations metrics and executive presentations.
Also, a note of caution. Beware IT or the CXO who “standardizes” on one tool. It’s like decreeing that all feet, regardless of size, must fit into one shoe of one size. No organization can get past the limitations of one technology and therefore a toolbox is needed.
If only data visualization is used then only the What and Where questions can be answered, but since decisions are about the future, the critical What Will and How To questions will be subjected to merely biased guesses and not the cool light of unbiased analytics.
The Workings of Analytics Tools
1. Blog BigML.com
2. Robert J Zwerling & Jesper H Sorensen, Finance Analytics Institute, Analytics Academy, AI & ML Basics
3. Caitlan Stanway-Williams, May 24, 2019, theinnovationenterprise
4. Ola Lutz, April 19, 2019, JPL-NASA, https://www.jpl.nasa.gov/edu/news/2019/4/19/how-scientists-captured-the-first-image-of-a-black-hole/
5. Dun & Bradstreet / Forbes Insights 2017 Enterprise Analytics Study
1. Dun & Bradstreet / Forbes Insights 2017 Enterprise Analytics Study
Not a member-scholar yet? Join our financial community here!
Identify your path to CFO success by taking our CFO Readiness Assessmentᵀᴹ.
For the most up to date and relevant accounting, finance, treasury and leadership headlines all in one place subscribe to The Balanced Digest.